Another Bungled CDC Study on Restaurants and Masks

Yes, the world may end tomorrow. Or the day after. Or in a billion years. The claim is not false, at least not for a few days, but it is trivial in the sense that it offers no set of causes or probability estimates. The problem isn’t that the earth ending tomorrow is as unlikely as the also not false claim that the New York Knicks may win the Stanley Cup someday, the problem is that the mechanisms by which either could occur are not carefully specified.  

If used to start panic, the claim that the world may end tomorrow is irresponsible to the point of possibly being self-fulfilling because if widely believed many people would commence pillaging and raping, or at least not working. Food and fuel would go undelivered, the Interlink would go down, and … you get the picture.

Most often, though, writers use “may” and “might” as weasel words. I know, because I have done it myself. If the “predicted” event comes to pass, the author can claim prescience. If it doesn’t, the author points to the hedge word so as to pretend not to have been wrong.

Sophisticated readers understand the “may” game but many readers apparently are not sophisticated enough to prevent news media from using it to sway readers toward editorial goals. So, for example, headlines blare “Easing restrictions may lead to an increase in Covid cases or deaths.” Right, and the media may be trying to manipulate people into pressuring leaders. Or maybe the reporter is a moron. Tough calls, all.

So I call not for cancelling “may” and “might” but rather for weighing probabilistic claims on the basis of evidence, logic, and theory. That the world may end tomorrow because a rock the size of Idaho is hurtling toward it and here are the mathematics of its speed and trajectory compared with that of this planet, the physics of gravity, and so forth is a much weightier proposition than the world may end tomorrow because it has to end sometime.

The same goes for Covid predictions. Arguing that lifting restrictions may cause a death spike because that is what happened under similar circumstances elsewhere is much weightier than the usual fare, which assumes that lockdowns reduce deaths and hence that ending them must increase them. 

The CDC is now citing a new study that shows that Covid “case” and death rates were lower in US counties with mask mandates and higher where on-site dining was allowed. The study is flawed because it doesn’t account for seasonality, position along the epi curve, the age or morbidity structure of the population, or lots of other things. Even with simplifying assumptions biasing the results in the desired direction, the difference the researchers find, even at the upper end of their estimate, is on the order of a few percent. Keep in mind that cases that don’t lead to hospitalization or death are good because they produce natural immunity. 

Any extra deaths are of course tragic so if Covid deaths were the only consideration this study would trigger the need for more research to parse out any actual causality between mask mandates (vs. voluntary masking) and restaurant closures (vs. individual adjustments like early bird dining for seniors only). While statistically significant (i.e., not likely random), the magnitude of the results suggests that masking and especially restaurant restrictions are much too costly, especially with vaccine rollouts continuing apace. It’s high time that bungling government public health bureaucrats stop placing most of the pandemic mitigation burden on service sector employers and employees.

Keep in mind that may work backwards in a sense too. Simply because X has not led to Y in the past does not mean that X may not lead to Y in the future, especially if X dramatically changes. So when I see The New York Times arguing that voter fraud will be minimal in the future because it was minimal in the past, I wonder if the editors of that once august newspaper are stupid, or if they simply believe that everyone else is? (I incline toward daft on Covid and condescending on election reform but have insufficient information to make the case.)

People weigh the expected costs and benefits of their behaviors, including criminal activities. If the expected benefit of the behavior increases and/or the expected costs decrease, then the prevalence of the activity will increase, regardless of past behavior. For example, nobody designed phone apps until the 2000s because the cost of doing so was very high and the benefit nil because nobody owned phones capable of running them. Now lots of people make phone apps because they are cheap to program and the markets for them are enormous.

Same thing goes for election fraud. Various states have changed election laws in ways that make fraud easier and cheaper while the increased growth of state power has palpably increased the benefits of fraud, electoral victory. So as Tom Fitton (esp. pps. 244-66), myself, and others predicted (though only as to sign, not degree), ballot harvesting/selling increased in 2020 and will increase yet more as people learn how to better engage in it (i.e., costs decrease) unless something is done to raise the cost of ballot fraud, either by rolling back some of the recent reforms like “Opt-out” Automatic Voter Registration and Universal Mail-In Ballots or developing new checks against abuses.

Or, we need to roll back the scope of government so that the benefits of electoral fraud decrease. I am not holding my breath on that one, though, until we figure out how to reduce humanity’s paternalistic tendencies.

* This article was originally published here

SHARE our articles and like our Facebook page and follow us on Twitter!

Post a Comment